1: \begin{abstract}

2: 

3: Full interpretation of data from gravitational wave observations

4: will require accurate numerical simulations of source systems, particularly

5: binary black hole mergers.

6: A leading approach to improving accuracy in numerical relativity

7: simulations of black hole systems is through fixed or adaptive mesh

8: refinement techniques.  We describe a manifestation of numerical interface 

9: truncation error which

10: appears as slowly converging, artificial reflections from refinement

11: boundaries in a broad class of mesh refinement implementations,

12: potentially compromising the effectiveness of mesh refinement techniques

13: for some numerical relativity applications if left untreated.  We elucidate this

14: numerical effect by presenting a model problem which exhibits the

15: phenomenon, but which is simple enough that its numerical error can be

16: understood analytically.  Our analysis shows that the effect is caused

17: by variations in finite differencing error generated across low and high

18: resolution regions, and that its slow convergence is caused by the

19: presence of dramatic speed differences among propagation modes typical of 

20: 3+1 relativity.

21: Lastly, we resolve the problem, presenting a class

22: of finite differencing stencil modifications, termed mesh-adapted differencing (MAD), 

23: which eliminate this pathology

24: in both our model problem and in numerical relativity examples.

25: 

26: 

27: %Numerically solving Einstein's field equations on a non-uniform mesh

28: %comes with its own unique set of challenges.  Here we investigate a

29: %particular type of noise that manifests whenever there is a non-zero

30: %shift vector across mesh interfaces.  By considering a model problem

31: %inspired by linearizing the Einstein equations, we discover the source

32: %of the noise and demonstrate a mechanism to remove it.  This method

33: %may have broad application.

34: 

35: \end{abstract}

36: